Predictive Analytics World for Government: Feedback

PAWgovI was at PAW Gov in Washington D.C. on September 12th and 13th and it was just great! Let’s start with the people. It was a pleasure for me to meet so many data mining experts. That was one great aspect of this PAW conference: experts are very accessible compared to other events. I had the opportunity to meet great people like Dean Abbott (Abbott Analytics), John Elder (Elder Research, co-author of the book Statistical Analysis and Data Mining), Eric Siegel (PAW Gov conference chair), Antonia de Medinaceli (Elder Research) and many more!

Continuing with the presentations, the focus was on fraud detection with other interesting topics such as tax evasion, crime analytics and postal applications. To be noted also keynotes by Earl E. Devaney and Congressman Darrell Issa.

I think PAW Gov was successful thanks to the following factors: many experts, exciting topics, unique track, narrow place (and thus friendly discussions during breaks). My only recommendation for next year is to have a few more technical presentations. However, the event was clearly successful and I recommend it without any hesitation!

Share

Recommended Reading

Comments Icon2 comments found on “Predictive Analytics World for Government: Feedback

  1. Hi Sandro, maybe it’s just my feeling, but lately the “big conferences” about datamining are more focused on advertising products and companies than methods, paradigms and experiences. I can imagine that no one wants disclose secrets algorithms and technologies, but my expectation in these kind of events is to exchange at least common thoughts on the common challenges.
    For example it would be interesting deepen more on “lessons learnt” about fraud detections, with practical exchange of different point of views on the open issues like: “how to create a good training set and what are the costs to do it”, “how to validate the strategies and the benefits of techniques deployed”.
    Otherwise the risk is to attend these events just to networking…
    http://textanddatamining.blogspot.com/

  2. @Cristian: Very good point! I would love to share and discuss publicly about the topics you mentionned. However, as you said, it is very difficult due to confidentiality and privacy reasons.

    The problem with fraud detection (and also with data mining in general) is that most of the application issues will be discovered in industry (not in academic world). And of course, industry is much less interested in sharing their ideas. In academic, you will not industrialize data mining and thus, you will never be confronted to the same issues (you can have your perfect data set, know which cases are frauds and have a clear evaluation strategy, etc.). This is much more difficult in industry.

    To come back to PAW Gov, I still think it was a nice event. However, it is true that it is more to network than to learn in the field (although there were really interesting presentations). Maybe an event such as KDD conference is much more technical.

    Feel free to share if you have any feedback about conferences!

Comments are closed.